摘要
视频分割是当前研究的热点问题.已有视频分割方法,只是考虑帧图像某点颜色特征,而没有考虑物体相邻像素相关性问题.提出了在基于GMM背景建模的基础上,同时引入像素点的邻域特征和像素点的色度和亮度特征对视频进行分割的方法.像素点的邻域特征能很好的解决因背景的微小变化而使分割效果较差的问题,色度和亮度特征很好的解决光照变化和阴影带来的问题.该方法先建立每个像素点的混合高斯模型,训练确定模型的相关参数,再结合像素点的领域特征,色度和亮度特征对视频进行分割.试验结果表明,该方法与其他方法相比在一定程度上改善了视频分割效果.
Video segmentation is a hot research issue. The color feature is only considered in the existing methods in video Segmentation. But the related issue among the object adjacent pixels is not considered. This paper proposed a new method of video segmentation based on GMM background modeling. In this paper, the adjacent feature of pixels, brightness and chromaticity of pixels are used to segment the video. The adjacent feature of pixels can be used to the problem that segmentation effect will change with the tiny variety of the background. The features of the brightness and chromaticity can be used to solve the problem of shadow. In this paper, the Gaussian mixture model for each pixel is built firstly. Then the values of parameters are trained. This model integrates with adjacent feature, brightness and chromaticity of the pixel to segment the video. Experiment results show that this method improve the performance of video segmentation comparing with other methods.
出处
《浙江工业大学学报》
CAS
2008年第1期81-85,共5页
Journal of Zhejiang University of Technology
关键词
混合高斯模型
视频分割
背景建模
gaussian mixture models
video segmentation
background modeling
